Comparison of morbidity and mortality after bloodstream infection with vancomycin-resistant versus -susceptible Enterococcus faecium: a nationwide cohort study in Denmark, 2010–2019

ABSTRACT

The emergence of bloodstream infections (BSI) caused by vancomycin-resistant Enterococci (VRE) has caused concern. Nonetheless, it remains unclear whether these types are associated with an excess risk of severe outcomes when compared with infections caused by vancomycin-susceptible Enterococci (VSE). This cohort study included hospitalized patients in Denmark with Enterococcus faecium-positive blood cultures collected between 2010 and 2019 identified in the Danish Microbiology Database. We estimated 30-day hazard ratio (HR) of death or discharge among VRE compared to VSE patients adjusted for age, sex, and comorbidity. The cohort included 6071 patients with E. faecium BSI (335 VRE, 5736 VSE) among whom VRE increased (2010–13, 2.6%; 2014–16, 6.3%; 2017–19; 9.4%). Mortality (HR 1.08, 95%CI 0.90–1.29; 126 VRE, 37.6%; 2223 VSE, 37.0%) or discharge (HR 0.89, 95%CI 0.75–1.06; 126 VRE, 37.6%; 2386 VSE, 41.6%) was not different between VRE and VSE except in 2014 (HR 1.87, 95% CI 1.18–2.96). There was no interaction between time from admission to BSI (1–2, 3–14, and >14 days) and HR of death (P = 0.14) or discharge (P = 0.45) after VRE compared to VSE, despite longer time for VRE patients (17 vs. 10 days for VSE, P < 0.0001). In conclusion, VRE BSI was not associated with excess morbidity and mortality. The excess mortality in 2014 only may be attributed to improved diagnostic- and patient-management practices after 2014, reducing time to appropriate antibiotic therapy. The high level of mortality after E. faecium BSI warrants further study.

KEYWORDS:

• Enterococci
• vancomycin
• antimicrobial resistance
• nosocomial
• morbidity
• mortality

Introduction

Enterococci are common causes of human infections. Enterococcus faecium and Enterococcus faecalis are associated with urinary tract infections, intra-abdominal infections, bloodstream infections (BSI), and endocarditis. In particular, E. faecium more often causes nosocomial infections, often related to intravenous catheters and other foreign devices [1–4]. In Europe, an increase in the occurrence of vancomycin-resistant Enterococci (VRE) has been observed, with E. faecium being the most dominant species [5,6]. Resistance to vancomycin can be encoded by several gene clusters with vanA and vanB genotypes being far the most prevalent among clinical VRE isolates [5]. From 2005 to 2019, Denmark observed an increase in VRE from clinical samples due to the emergence of several E. faecium vanA clones [7,8]. In addition, from 2016 and onwards, there was a national spread of a vancomycin-variable enterococci (VVE) clone, ST1421-CT1134 vanA E. faecium, i.e. a clone having unclear/variable vancomycin susceptibility but treated clinically as VRE [8,9]. In contrast to the increase in vanA E. faecium (VRE), the incidence of vancomycin-resistant E. faecalis (vanA or vanB) has remained at a stable low level [7].

In 2009, the 30-day mortality among Danish patients with E. faecium BSI was found to be very high compared with the mortality for E. faecalis BSI patients (38% vs. 20%) [10]. International meta-analyses [11–14], as well as later studies [15], including a combined case series from Denmark and the Netherlands [16], further suggest an association between VRE BSI and excess mortality and length of hospital stay. However, methodological biases such as patient selection and unmeasured confounding factors make interpretation difficult. To investigate this further, we used un-selected nationwide microbiological surveillance data to establish a large cohort of all E. faecium BSI patients from 2010 to 19 in Denmark, to estimate 30-day mortality and discharge rates for VRE BSI compared with VSE BSI, while excluding VVE cases and adjusting for age, sex, comorbidity, length of stay and other factors.

Methods

Data sources

The Danish Civil Registration System (CRS) contains continually updated demographic variables and the vital status of all Danish citizens since 1968. A unique personal 10-digit identification number permits tracking of information over time and accurate linkage between national registers and databases [17]. The Danish Microbiology DataBase (MiBa) serves both as a tool for clinical doctors to access clinical microbiology test results on their patients as well as a national surveillance and research database. MiBa was established in January 2010 in a collaboration between the 10 Departments of Clinical Microbiology (DCM) in Denmark, and the National Reference Laboratories (NRL) at Statens Serum Institut [18,19]. The DCMs serve all hospitals, general practitioners, and clinical specialists in Denmark, and data from microbiological analyses performed by the DCMs are registered in MiBa. Each test report in MiBa contains information on the individual case (personal identification number, age at sample, sex), the sample (sample ID, sampling date and time), specimen type (blood, urine, etc), the diagnostic analyses performed, and results including microbiological species, phenotypic antimicrobial susceptibility and in some cases resistance genes (e.g. vanA and vanB) [18–20]. The DCMs also submit samples directly to NRL for species identification, genotyping, and surveillance purposes. The results are however not registered in MiBa, yet, but in separate databases at SSI. Samples must be submitted for notifiable diseases but are voluntarily also submitted for other important diseases and infections, e.g. VRE. Thus, since 2005, VRE isolates from clinical samples (i.e. not including isolates from faecal screenings) have been submitted voluntarily by DCMs [7,8]. Genotyping was done for vanA and vanB by PCR from 2005 through 2014, and from 2015 through Q1 2019 all clinical VRE/VVE isolates underwent whole-genome sequencing (WGS) as previously described [7,8]. This data was used elsewhere to describe the increase in patients with VVE (ST1421-CT1134 vanA) isolated from all types of clinical samples submitted (not only blood) (2015, 0/375 isolates, 0%; 2016, 1/435 isolates, <1%; 2017, 12/426 isolates 3%; 2018, 167/518 isolates, 32%; 2019, 285/584 isolates, 49%) [8,21]. The National Patient Register (NPR) contains information from all public and private hospitals about diagnoses and procedures, for inpatients since 1977 and outpatients (ambulatory) since 1995, as well as information on unit, specialty of the unit, and date and time of admittance and discharge for each contact to a unit [22]. As described in detail elsewhere, the duration of a hospital stay (or a “course”) is calculated as the time from admission to the first unit until discharge from the last unit [23]. Since 1994, diagnoses have been coded using version 10 of the International Classification of Diseases (ICD).

Ethics

This article has been prepared on the basis of a study carried out as part of a task imposed on Statens Serum Institut according to the national legislation as specified in the Danish Health Care Act (Sundhedsloven § 222). The need for ethics approval and informed consent is therefore deemed unnecessary according to national legislation, cf. implementing decree 2020-09-01, number 1338, about the scientific regulatory procedure of health science research projects and health data scientific research projects. The article only contains aggregated results and no personal data. The article is therefore not covered by the European General Data Protection Regulation.

Definitions of VRE, VSE, and VVE

Only VRE and VSE BSI with E. faecium were included as exposures, while VVE BSI with E. faecium was excluded and occurred mainly in 2018 and 2019 and later as also described elsewhere [8,9,21]. Blood isolates of E. faecium were defined as a VRE if the DCM reported them vancomycin-resistant in MiBa and isolates were defined as VSE if the DCM reported them vancomycin-susceptible in MiBa. The methods for detection of VRE, VSE, and VVE at the 10 individual DCMs have varied both between the DCMs and in the study time period. All DCMs have used a combination of agar diffusion methodology using breakpoints and rules defined by either The Clinical & Laboratory Standards Institute (CLSI) or The European Committee on Antimicrobial Susceptibility Testing (EUCAST) and PCR setups detecting vanA/B genes [21,24]. All DCMs have also in the entire study period participated in national and international quality assurance collaborations such as the United Kingdom National External Quality Assessment Schemes (NEQAS).

We excluded patients with a VVE BSI with E. faecium defined in two ways in order to use all the available national data to identify as many as possible of these patients: (1) By using data from MiBa, i.e. data from the DCMs reported just as for VRE and VSE, and (2) by using data from prior WGS of the VVE clone ST1421-CT1134 vanA in all isolates submitted by DCMs to NRL at SSI [8], and restricted to blood isolates collected from patients in the study period. The first way defined a proxy for VVE, where we took advantage of the unique characteristic that the VVE clone may be cultured vancomycin-susceptible even though it has the vanA gene. Thus, we excluded patients with blood isolates for which the DCM had registered lab results in MiBa (i.e. through a MiBa data-transfer protocol) showing that the isolates were cultured phenotypic vancomycin-susceptible based on MIC/zone data and having E. faecium vanA gene identified by PCR. In the second way, we used a dataset of 64 patients identified separately with the ST1421-CT1134 vanA E. faecium clone in submitted E. faecium blood isolates, to exclude further VVE patients from the study population, if not already excluded. This second way of exclusion was introduced to the study later in June 2023 and used to perform additional analysis to check coherence with results using the first way, introduced to the study in 2019.

Study design

The study was an observational cohort study including two exposure groups, patients with E. faecium VRE or VSE BSI detected during a hospital stay and then followed up for the outcomes of discharge or death within 30 days. The follow-up time was designed so that a patient could change exposure status during the follow-up time. For example, if the first collected blood sample was positive for VSE and the second one was positive for VRE, the patient would change status from VSE to VRE, while the time from the first to second sample still counted as follow-up time. In this way, the final VSE group could not include any VRE patients, and the start of follow-up could not be biased by future exposure status.

Study cohort of patients

Figure 1 shows a flow diagram of the inclusion criteria of the study cohort. Thus, the source population consisted of 6317 patients identified in MiBa with 10,903 E. faecium-positive blood cultures collected between 01 January 2010, and 29 November 2019. The 10,903 E. faecium blood isolates were marked as either VRE, VVE, or VSE isolates using the definitions described further above. We first excluded VVE patients by definition “way 1” (i.e. using MiBa data only), and then isolated from blood cultures or blood culture sets collected at the same time, leaving 8687 isolates with VRE or VSE eligible for inclusion. Next, by linkage with the National Patient Registry, 6102 of the 6272 patients had 8414 of the 8687 sample set time points linked to the period of their hospital stay(s). The remaining 170 (3%) patients were excluded because sample time points could not be placed in any hospital stay, likely due to ongoing admissions and registration delay for admissions at the time of data extraction. Next, only the first VRE sample set was included. The first VSE sample set was included when it occurred before or without, but not after an eventual VRE sample set in the same patient. In this way 6153 of the 8414 samples were left for inclusion from the 6102 patients. Next, 31 patients were excluded due to less than 30 days of follow-up from the first E. faecium BSI sample time point until the end of follow-up on 3 December 2019, leaving 6071 patients with 6111 sample time points eligible for inclusion in the patient study cohort. For an additional analysis, we also excluded a further 54 VVE patients identified by definition way 2 (i.e. using WGS data) described further above.

Figure 1. Study flow diagram.

Co-variates

Charlson Comorbidity Index (CCI) score was based on Quan et al. (17 disease groups) [25], and calculated based on the latest 5 years of diagnoses before the blood sample time point, as obtained by linkage with the National Patient Registry using the unique personal identification number and date-time interval for the start and end contact of each diagnosis. Intensive care unit (ICU) contact was defined using procedure codes and dates for intensive therapy/observation (NABB, NABE), respirator ventilation (BGDA0), non-invasive ventilation (BGDA1), acute dialysis treatment (BJFD0), and treatment with inotropica/vasopressors (BFHC92, BFHC93, BFHC95) as described elsewhere [26,27]. Other co-variates such as age, year, etc., are evident or described in the “Data sources” section.

Statistical methods

The effect of VRE BSI compared to VSE BSI on the risk of death or discharge to home was evaluated using hazard ratios (HR) from a Cox proportional hazards regression model with time since the first E. faecium BSI sample as the underlying time scale. The outcomes of discharge or death were used as strata. The discharge outcome was analyzed with censoring at the first occurrence of either discharge, death or 30 days after the first VRE/VSE sample. For the death outcome, all patients were censored at the first occurrence of either death or 30 days after the first VRE/VSE. The model was adjusted for sex, age at the sample time point (10-year groups), and CCI (9 groups). The interaction between time from admission until the first E. faecium BSI sample (1–2, 3–14, and >14 days) and the effect of VRE vs. VSE on the HR of death or discharge to home was evaluated by including an interaction term in the model. Other risk factors, e.g. for the specialty of unit, region of hospital, and calendar year of E. faecium BSI, were evaluated by including interaction terms in the model. In a sensitivity analysis, the endpoint of death was restricted to in-hospital mortality (i.e. excluding death at home). We evaluated differences in proportions of characteristics at entry using χ2 test for categorical data, approximate Wilcoxon Rank Sum Test for continuous data, and for cases per year also the Cochran–Armitage trend test. All analyses were performed using the SAS statistical software version 9.4 (SAS, Cary Inc., USA), e.g. PROC PHREG procedure for the Cox regression.

Results

The study cohort included 6071 patients with E. faecium BSI detected after a median of 10 days of hospital stay (IQR, 2–21 days), 335 in the VRE group and 5736 in the VSE group.

Characteristics of the VRE and VSE group

Compared with the VSE group, the VRE group included less elderly (median 68 vs. 70 years for VSE), more patients from the Capital Region (which in general has a higher hospital activity (bed-days) than other regions [21] and thus a higher risk of outbreaks with nosocomial infections), and increased during the study period (Table 1). A slightly lower proportion of VRE patients had prior hospital contacts than among VSE patients (95.8% vs. 97.4%), but the VRE patients had more recent contacts (median 4 vs. 6 days since last contact) and a higher number of contacts (40 vs. 32 contacts), than the VSE patients. In general, there were few outpatients in the study cohort but slightly more in the VRE group (2.4% vs. 0.9%). The VRE group included patients with longer stays (median 35, vs. 28 days for VSE), which was due to longer pre-BSI stay, while the post-BSI stay (time to discharge or death) did not differ significantly between the two groups, i.e. the cause of longer VRE stays was a later occurrence of VRE than VSE rather than a longer duration of stay after the VRE BSI. In general, comorbidity was high but similar between groups with a median Charlson comorbidity index of 2 and a high proportion of patients with cancer (43.3% vs. 43.3%). The proportion of patients with renal disease was higher in the VRE group than in the VSE group (17.9% vs. 13.7%) (see eTable 1 for all groups). The proportion treated in an intensive care unit was high but not different between the VRE and VSE groups (44.5% vs. 44.4%) (Table 1). VRE patients were typically diagnosed in units or departments with a specialty in anaesthesia, haematology, nephrology, gastroenterology medicine, and surgical gastroenterology (55.7% vs. 40.5% for VSE) (Table 1 and eTable 2).

Table 1. Characteristics of 6071 hospital patients at the time of bloodstream infection with vancomycin-resistant or -susceptible Enterococcus faecium, all hospitals in Denmark, 2010–2019.

Characteristic at the time of E. faecium BSI	VRE (N = 335)	VSE (N = 5736)	P-values
Age, years, median (IQR)	68 (58–76)	70 (61–77)	0.006
Males, n (%)	200 (59.7)	3488 (60.8)	0.69
Region of hospital, n (%)
Capital	205 (61.2)	2085 (36.3)	<.0001
Central Denmark	62 (18.5)	1001 (17.5)	.
North Denmark	9 (2.7)	434 (7.6)	.
Zealand	33 (9.9)	668 (11.6)	.
Southern Denmark	26 (7.8)	1548 (27.0)	.
Year, n (%)
2010	10 (3.0)	471 (8.2)	<.0001
2011	8 (2.4)	545 (9.5)	.
2012	12 (3.6)	569 (9.9)	.
2013	26 (7.8)	568 (9.9)	.
2014	33 (9.9)	658 (11.5)	.
2015	36 (10.7)	613 (10.7)	.
2016	47 (14.0)	569 (9.9)	.
2017	52 (15.5)	653 (11.4)	.
2018	67 (20.0)	648 (11.3)	.
2019	44 (13.1)	442 (7.7)	.
Comorbidity (latest 5 years),
Charlson comorbidity index, median (IQR)	2 (1–4)	2 (1–4)	0.33
Charlson comorbidity groups, n (%)
Renal disease	60 (17.9)	787 (13.7)	0.03
Cancer	145 (43.3)	2481 (43.3)	0.99
Metastatic carcinoma	29 (8.7)	487 (8.5)	0.92
Chronic pulmonary disease	50 (14.9)	1039 (18.1)	0.14
Diabetes without complications	61 (18.2)	1111 (19.4)	0.60
Other groups (see Table 1)	47 (14.0)	971 (16.9)	0.38
No mapping of comorbidity	49 (14.6)	788 (13.7)	0.65
Prior hospital contacts (latest 5 years)
Yes, n (%)	321 (95.8)	5587 (97.4)	0.08
Contacts to units*, n, median (IQR)	40 (15–78)	32 (13–67)	0.0007
Combined duration*, days, median (IQR)	30 (7–66)	27 (8–60)	0.12
Time since last contact, days, median (IQR)	4 (1–17)	6 (2–27)	0.07
Admission type, n (%)
Outpatient (stays <12 h)	8 (2.4)	54 (0.9)	0.01
Inpatient (stays >12 h)	327 (97.6)	5682 (99.1)	.
Duration of hospital stay, median (IQR), days
Total length of stay	35 (21–65)	28 (14–52)	<.0001
Since admission (using 1st VRE in VRE group)	17 (6–30)	10 (2–20)	<.0001
Since admission (using 1st VSE in VRE group)	17 (5–29)	10 (2–20)	<.0001
Until discharge to home within 30 days	13 (8–21)	12 (8–19)	0.26
Until death within 30 days	6 (3–13)	7 (3–14)	0.85
No outcome in hospital within 30 days**	54 (39–82)	51 (39–76)	0.68
ICU since admission, n (%)	149 (44.5)	2548 (44.4)	0.98
Specialty of unit, n (%)
Anaesthesiology	48 (14.3)	534 (9.3)	<.0001
Haematology	52 (15.5)	686 (12.0)	.
Nephrology	14 (4.2)	118 (2.1)	.
Gastroenterology medicine	20 (6.0)	222 (3.9)	.
Surgery gastroenterology	46 (13.7)	762 (13.3)	.
Other specialties (see Table 2)	155 (46.3)	3414 (59.5)	.

VRE, vancomycin-resistent E. faecium; VSE, vancomycin-susceptible E. faecium; BSI, bloodstream infection; BC, blood culture; IQR, interquartile range; ICU, intensive care unit.

*A hospital stay may consist of multiple contacts to units.

**Length of stay until discharge or death more than 30 days later.

Outcome after VRE vs. VSE

The 30-day mortality rate was 37.6% for VRE (126 of 335 patients) and 37.0% for VSE (2123 of 5736 patients), while the 30-day rate of discharge for the corresponding groups was 37.6% for VRE (126 of 335 patients) and 41.6% for VSE (2386 of 5736 patients), leaving 24.8% for VRE (83 of 335) and 21.4% for VSE (1227 of 5736) still in hospital more than 30 days after the BSI.

Table 2 shows the main analyses of discharge and mortality HRs for VRE vs. VSE in the study cohort of 6017 patients. Overall, neither crude HRs, nor HRs adjusted for sex, age, and comorbidity (hence forward only adjusted HRs and p-values are shown in the Results section), were increased for death (HR 1.08, 95%CI 0.90–1.29) or reduced for discharge (HR 0.89, 95%CI 0.75–1.06), when comparing the VRE with the VSE group. Results were very similar when using age as a continuous adjustment variable rather than as a categorical variable in 10-year groups (data not shown).

Table 2. Discharge and mortality hazard ratios for vancomycin-resistant versus susceptible Enterococcus faecium bloodstream infection, 6071 patients (335 VRE, 5736 VSE), Denmark, 2010–2019.

Exposure	Outcome
	Discharge to home (within 30 days after BSI)		Death (within 30 days after BSI)
	Unadjusted hazard ratio	Adjusted hazard ratio	Unadjusted hazard ratio	Adjusted hazard ratio
	(95% CI)	(95% CI)	(95% CI)	(95% CI)
VSE	1 (ref.)	1 (ref.)	1 (ref.)	1 (ref.)
VRE	0.87 (0.73–1.04)	0.89 (0.75–1.06)	1.06 (0.88–1.26)	1.08 (0.90–1.29)
	P = 0.13	P = 0.21	P = 0.55	P = 0.41

VRE, vancomycin-resistant E. faecium; VSE, vancomycin-susceptible E. faecium; BSI, bloodstream infection. Hazard ratios are adjusted for age (10-year groups), sex, and Charlson comorbidity (9 groups); CI, confidence intervals.

Because Table 1 did not suggest a longer and thus more severe course of BSI with VRE than VSE (i.e. longer VREs stays appeared to be due to longer pre-BSI stay – and not post-BSI stay), we examined the interaction between time since admission (grouped as 1–2, 3–14, and >14 days) and VRE vs. VSE on the HR of death and discharge. Overall, there was no interaction, neither for death (P_interaction= 0.14) nor discharge (P_interaction= 0.45) (Table 3).

Table 3. Discharge and mortality hazard ratios for patients with vancomycinresistant versus -susceptible Enterococcus faecium bloodstream infection by time since admission, 6071 patients (335 VRE, 5736 VSE), Denmark, 2010–2019.

Exposure	Outcome
	Discharge to home (within 30 days after BSI)		Death (within 30 days after BSI)
	Unadjusted hazard ratio	Adjusted hazard ratio	Unadjusted hazard ratio	Adjusted hazard ratio
	(95% CI)	(95% CI)	(95% CI)	(95% CI)
By time since admission
1–2 days (n = 1496)
VSE	1 (ref.)	1 (ref.)	1 (ref.)	1 (ref.)
VRE	0.81 (0.56–1.16)	0.80 (0.55–1.17)	0.85 (0.51–1.40)	0.80 (0.49–1.30)
3–14 days (n = 2333)
VSE	1 (ref.)	1 (ref.)	1 (ref.)	1 (ref.)
VRE	1.09 (0.80–1.50)	1.08 (0.79–1.48)	0.92 (0.65–1.31)	0.90 (0.64–1.29)
>14 days (n = 2287)
VSE	1 (ref.)	1 (ref.)	1 (ref.)	1 (ref.)
VRE	1.04 (0.79–1.37)	1.04 (0.79–1.37)	1.17 (0.93–1.46)	1.24 (0.99–1.56)
	P = 0.64	P = 0.45	P = 0.50	P = 0.14

P-values are for interaction terms; VRE, vancomycin-resistant E. faecium; VVE, vancomycin-variable E. faecium; VSE, vancomycin-susceptible E. faecium; BSI, bloodstream infection. Hazard ratios are adjusted for age (10-year groups), sex, and Charlson comorbidity (9 groups); CI, confidence intervals.

The results also showed that e.g. region, year, and medical specialty were different between the VRE and VSE groups (Table 1). We therefore performed interaction analyses (eTable 3) to reassure the main finding of no excess VRE mortality. First, for the region of hospitals in which the E. faecium BSI was diagnosed, there was no interaction with the effect of VRE vs. VSE on the HR of death (P_interaction= 0.38) or discharge (P_interaction= 0.65). Next, we also found no overall interaction between the calendar year of E. faecium BSI (2010–2019, 1-year categories) and VRE vs. VSE on the HR of death (P_interaction= 0.40) or discharge (P_interaction= 0.06). However, there was an 87% excess mortality in 2014 for VRE vs. VSE (HR 1.87, 95%CI 1.18–2.96) and a 53% excess discharge in 2019 (HR 1.53, 95%CI 1.05–2.22). eFigure 1 shows the excess by year and indicates a standstill in VRE excess mortality and VRE discharge reductions from 2015 to 2019. Furthermore, there was no interaction between the medical specialty of the unit in which the E. faecium BSI was diagnosed (grouped as shown in Table 1 and defined in eTable 2) and VRE vs. VSE on the HR of death (P_interaction= 0.07) or discharge (P_interaction= 0.81). However, there was a 95% excess mortality for VRE patients within the anaesthesiology specialty (HR 1.95, 95%CI 1.28–2.96). More than 80% of the patients in this specialty were admitted in units/departments of anaesthesiology/surgery (33%), anaesthesiology (9%), or intensive/semi-intensive care beds (39%) at the time of the blood sample identifying VRE and VSE (data not shown).

In a sensitivity analysis, we repeated the main analysis and excluded VRE patients if they had VSE before the VRE, and the findings were compatible with the main finding (death, HR 0.98, 95%CI 0.82–1.17; discharge, HR 0.93, 95%CI 0.78–1.11); 24 patients were excluded, who constituted 7% of the VRE group and had a 30-day mortality rate of 37.5% (9 of 24) and a discharge rate of 16.7% (4 of 24). In another sensitivity analysis, we restricted the endpoint to in-hospital mortality (excluding mortality after discharge) and the result was similar to the main analysis (death, HR 1.08, 95%CI 0.89–1.30).

Figure 2 shows the crude cumulative incidence of death and discharge to home according to days after VRE and VSE (i.e. without adjustment for age, sex, and comorbidity). The figure indicates that from day 7 to 30, VRE patients had small excess mortality of <3%, and <5% fewer patients were discharged to home when compared with the VSE group. However, these differences were insignificant according to the overall 30-day HR estimates in both the crude and adjusted analyses (Table 2).

Figure 2. Cumulative incidence of patients who died within 30 days (top figure) or were discharged (bottom figure) after vancomycin-resistant (VRE), and -susceptible (VSE) Enterococcus faecium bloodstream infection, 6071 patients, all hospitals in Denmark, 2010–2019.

In an additional analysis, we excluded a further 54 VVE patients from the study population using WGS data (way 2, see Methods section and Figure 1) on the ST1421-CT1134 clone, identified in 64 patients of whom 10 were already excluded using MiBa data (way 1). As in the main analysis, the overall HR for death (HR 1.09, 95%CI 0.89–1.34) was not different between the VRE and VSE groups, while the overall HR for VRE discharge was reduced (HR 0.81, 95%CI 0.66–0.98) but numbers were now smaller, specifically in 2018 and 2019 (due to exclusion of the VVE cases) where most VRE cases in the study occurred and discharge was high (See eFigure 1). Among all 99 (45 + 54) VVE patients excluded by way 1 + 2, the crude 30-day mortality was 36.6% in 2018–2019, compared with 37.6% for VSE and 36.9% for VRE in 2018–19.

Discussion

This nationwide study did not find an excess mortality or admission burden after E. faecium VRE compared to VSE BSI in Denmark in the entire study period 2010–2019. Together with the spread of VRE, which started in this period, then followed an excess mortality in 2014 and then a standstill in mortality from 2015 to 2019, which we suggest could be due to better diagnostics and screenings, reducing the time to appropriate antibiotic treatment for VRE. For results in 2018–2019, a bias cannot be completely ruled out with the available data on the VVE clone, as explained further below (see third last section).

Previous analysis of Danish data is limited to one study, which reported an excess 30-day mortality among 27 Danish VRE E. faecium BSI patients from four hospitals in the Capital Region in 2012–2014 pooled with 36 cases from the Netherlands (overall HR 1.54, 95%CI 1.06–2.25) [16]. The comparison group included 104 Danish ampicillin-resistant E. faecium BSI cases, which parallels the E. faecium VSE BSI cases in the present study as nearly all E. faecium cases are ampicillin-resistant. Notably this comparison group had similar 30-day mortality as the comparison group in the present study (38%) – and also as in an earlier Danish study of a similar group from 2006 to 2009 [10]. However, for the 27 Danish E. faecium VRE BSI they found a higher 30-day mortality of 48%, vs. 38% for the parallel group in the present study. We speculate that explanations for this difference could be partly related to differences in diagnostic- and patient-management practices in the different study periods, i.e. 2009–14 vs. 2010–2019. In the present study, most VRE cases were from 2015 to 2019 (73%) and in that period two circumstances have improved. Firstly, direct identification from positive blood culture bottles with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MaldiToF) was introduced, thus giving species identification on the same day as the blood culture turns positive and thereby the possibility of initiating E. faecium-directed therapy with vancomycin or with linezolid/daptomycin in case VRE could be expected in the individual patient. One DCM in the Capital Region (where most VRE are diagnosed) performs a subsequent direct vanA/B PCR on E. faecium-positive bottles. This workflow gives the diagnosis of VRE on the same day as the blood culture turns positive and enables prompt appropriate antibiotic therapy. Although species identification on the same day by MaldiToF in this way could help explain why we observed no excess VRE BSI mortality after 2014, it is however not entirely compatible with the constant high mortality for VSE BSI that we observed all years. Last but not least, attention and screening for VRE increased with VRE cases. In general, all peer-patients to an index case of VRE (e.g. detected in urinary isolate randomly) were screened and the index case isolated, but as VRE increased, patients were also screened e.g. after travel abroad, after readmission within 6 months (then also isolation occurred), and when transferred from units with many outbreaks, et cetera. In this way, VRE colonization was captured earlier and the time to appropriate antibiotic therapy if BSI later occurred was reduced. These two explanations for the difference between ours and the Danish-Netherlandish study are supported by our observations of a tendency for VRE discharge rate reductions in 2012–2014, and furthermore an 87% excess VRE mortality in 2014 only, and even a 53% excess in VRE discharge rates later in 2019. The large size of the present study made it possible to observe this standstill in excess burden despite increases in VRE, corroborating the importance of improving practices, in particular when effective therapy was, and still is, available, e.g. linezolid and daptomycin. The hospitals’ use of linezolid and daptomycin has increased in the study period (see eFigure 2) [21] together with increases in VRE. In addition, few cases with linezolid-resistant and linezolid-vancomycin resistant enterococci have been observed in the study period – although their presence is worrying [21]. Interestingly, in contrast to our study the Danish-Netherlandish study had information on individual antibiotic use from clinical records, and also reported that time to appropriate antibiotic therapy could not explain the increased mortality in VRE bacteraemia they observed. The authors defined appropriate therapy for VRE E. faecium BSI as the use of linezolid, daptomycin, and/or tigecycline, however, noted that, in the Danish data, this was often a combination treatment with linezolid and daptomycin. Although these effect of time estimates were uncertain, it was evident that the overall effect, i.e. when not considering intervals of time to appropriate therapy, was that VRE BSI patients on inappropriate therapy did fare worse than the reference group of ampicillin-resistant E. faecium BSI patients on inappropriate therapy. The authors suggested that the excess VRE mortality they observed could be due to increased virulence of VRE or unmeasured confounding [16].

Studies from other countries also to some degree contradict our findings for Denmark. Thus, a meta-analysis by Prematunge et al., of twelve cohort studies from 1997 to 2014 in eight countries, concluded that VRE BSI (n = 643 vs. n = 1970 VSE BSI) remains associated with excess mortality (odds ratio 1.80, 95% CI 1.40–2.32) and an excess total length of hospital stay (mean 5 days), despite the availability of effective VRE therapy since 1999 (quinupristin-dalfopristin, linezolid, or off-label daptomycin) [13]. The excess was consistent with two other meta-analyses including studies prior to effective VRE therapy (risk ratio 2.38, 95% CI 2.13–2.66 and odds ratio 2.52, 95% CI 1.87–3.39) [11,12]. A more recent meta-analysis by Brinkwirth et al. of studies from 2010 to 2020 in five European countries reported a higher pooled mortality among VRE BSI patients from two studies than in general among Enterococcus species BSI patients from five studies (32.5% vs. 21.9%) [14] but also emphasized an Irish study of causative organisms in BSI that reported moderate attributable mortality for both E. faecium VRE (19.1%) and VSE (17.3%) [28]. Finally, a recent cohort study of enterococcal BSI in 11 US hospitals in 2016–2018 concluded that the impact of VRE on mortality for 56 VRE vs. 176 VSE patients changed during the hospital stay [15]. With these findings, it still remains debated whether VRE BSI results in higher mortality than VSE BSI. Several biases were scrutinized also in the meta-analyses. Publication bias was out ruled in the pre-VRE therapy meta-analysis [12], and Brinkwirth et al. did not report on this [14]. Prematunge et al. observed no logical publication bias (i.e. it was driven by two small cohort studies with insignificant estimates below unity) and put emphasis on bias from unadjusted differences (i.e. few studies adjusted for confounders), differences in treatment effectiveness, and importantly differences in proportions of patients with E. faecalis and E. faecium in the VRE and VSE group. In the present Danish setting, we however observed neither unadjusted or adjusted overall excess 30-day mortality for E. faecium VRE, and furthermore the 30-day discharge rate for VRE were not reduced. In contrast to the majority of meta-analyzed studies, our study was restricted to E. faecium BSI patients, and E. faecalis BSI rarely co-occurred (ie. in approximately 4%, with no difference between the VRE and VSE group; data not shown).

In line with the meta-analysis by Prematunge et al., we also observed an excess total length of stay for VRE BSI patients, roughly an excess of one week [13]. However, we observed this for the pre-BSI length of stay only (median, 17-10 = 7 days), rather than the post-BSI length of stay as might be expected if VRE BSI was to be associated with more severe outcomes than VSE BSI. Similarly, in the meta-analysis, there was no excess for post-BSI length of stay, and it was not possible to address pre-BSI length of stay [13]. Thus, our findings add novel data on pre-BSI length of stay and support that the longer hospital stays in VRE patients are mainly due to the time elapsed in the hospital before acquiring a VRE BSI. The longer time to VRE than VSE BSI can be ascribed to the fact that VRE infections are health care-associated and rarer than VSE, and hence, the risk of infection increases only after a longer period in hospital. In the adjusted analyses, we stratified mortality and discharge HRs by period in hospital, and reassuringly, we also did not observe significant excess mortality or delayed discharge even when VRE BSI occurred 14 days or more after admission.

We also observed that E. faecium VRE BSI patients were more common in hospital departments/units with a specialty in anaesthesiology, haematology, nephrology, gastroenterology medicine, and surgical gastroenterology than in other specialties. In interaction analysis, there was excess mortality for VRE patients specifically within the anaesthesiology specialty that in turn could be connected mainly to ICU bed units and surgery. It is noted that the mortality rate in this study was higher for VSE BSI than found in other studies [15,28,29]. In addition, for E. faecium BSI in general, the presented data vs. recent results from analysis of data on blood isolates from the European Antimicrobial Resistance Surveillance Network (EARS-Net) 2012–2018, respectively, together indicate that E. faecium BSI is more common in ICU (44% vs. 25%), internal medicine units (16% vs. 23%), and surgical units (31% vs. 11%), than E. faecalis (percentages not shown for E. faecalis) [6]. The even higher occurrence in ICU and surgical units in the present data vs. EARS-Net data might offer some clue to the high overall mortality we observed for E. faecium VSE BSI, as well as for E. faecium BSI in general, but could however, also reflect differences in e.g. registration and patient populations. Recently, a study of 599 VSE E. faecium BSI patients in one Danish hospital, reported that although the 30-day mortality was 40%, clinical data showed that only in 6.3% of cases was VSE E. faecium BSI the likely cause of death [30]. Thus, the causality of 30-day mortality should probably be interpreted with caution, also for antibiotic-resistant E. faecium BSI.

During the study period, a VVE clone emerged (ST1421-CT1134 vanA E. faecium) in Denmark, which we excluded due to its variable susceptibility to vancomycin, i.e. it may be resistant or sensitive to vancomycin while having the vanA gene. As a precaution, it is treated clinically as a VRE in the hospitals. The clone was initially detected in 2016 by DCM at Hvidovre Hospital in Denmark [9]. Based on all the 10 Danish DCMs’ voluntary submission of clinical isolates, the NRL at SSI documented an increase in ST1421-CT1134 vanA E. faecium, isolated from all types of clinical samples since 2016, and from blood cultures since 2018 [8,21]. In support of the later, Hvidovre Hospital, performing the initial detection, also detected only 5 isolates with ST1421 among all E. faecium blood culture isolates (n = 169) from patients in 2016 and 2017 (first identifying vanA by PCR, then ST1421 by sequencing) [9]. To obtain the most complete exclusion of VVE BSI patients in 2018 and 2019, we first used MiBa data to define a proxy for vancomycin-sensitive VVE, and next used our WGS data from NRL at SSI. The crude 30-day mortality in the entire excluded VVE group did not suggest mortality was different from the included VRE and VSE groups. However, we cannot rule out some VVE BSI patients were not excluded and that this to some degree could have biased the results for 2018 and 2019 towards no VRE burden. It might be argued that other clonality than VVE could explain our null-finding, e.g. due to dominance of a specific virulent clone of VSE causing the high mortality, or dominance of a less virulent clone of VRE causing lower mortality. However, genotyping data suggests single clones are unlikely to dominate. For example, for VSE, a recent study from Odense University hospital in region Southern Denmark showed that E. faecium VSE blood isolates collected between 2015 and 2019 displayed 42 sequence types (STs) and 131 complex types (CTs) in several clusters, based on 630 isolates from 599 patients [30]. Thus, the results do not indicate our null finding was due to a single or few high-virulent E. faecium VSE clones. For VRE, NRL’s routine genotyping of isolates from hospitals since 2015 shows that clonality is relatively high with successive single clones dominating in different time periods [21] (DANMAP report 2022, p. 139). Thus, we believe it is unlikely that a single low-virulent VRE clone could explain our null finding.

The strengths of the present study include the use of national surveillance data over a 10-year period, a large unselected patient cohort with VRE/VSE BSI, and adjustment for confounders. The largest previous study included 116 VRE BSI patients [31] and the two most recent meta-analyses included a total of 650 VRE BSI patients [13,14] – compared with 335 VRE BSI patients in the present single study. Many previous studies included selected patient groups (e.g. stem cell transplantations [32–34], patients with hospital stays >2 days [31], tertiary care patients [35]) and such studies may therefore also select for a higher mortality. The observational nature of the present study means e.g. that the number of blood samples per patient depended on the diagnostic build up for each patient – and not a predesigned number of blood samples for all patients e.g. as preferred in a clinical trial or prospective study. This might potentially bias the estimate of time from E. faecium BSI until death or discharge. However, a strength of the present study was that such potential bias was reduced by starting follow-up at the collection of the first E. faecium BSI isolates regardless of VRE/VSE status. The use of national surveillance data was also the main limitation of our study, because of the absence of specific individual –level data, e.g. on the clonal relationship of the included isolates as well as clinical focus of the BSI, the cause of death, and data on other infections (except E. faecalis) such as in polymicrobiology BSI or as detected in other types of isolates than blood. Also the absence of data on antibiotic treatment of the BSI could possibly bias results but we would expect that to favour an increased burden of VRE caused by delayed onset of appropriate therapy (although this was previously investigated [16] as discussed further above). Optimization of therapeutics for VRE BSI is furthermore complex, including not only factors such as time to initiation of antibiotics, but also e.g. appropriate dosing, pharmacokinetics, and pharmacodynamics.

In conclusion, VRE E. faecium BSI in Danish hospitals was not associated with excess 30-day morbidity and mortality, regardless of time since admission, when compared to VSE E. faecium BSI. However, the observed increase in VRE BSI is a concern because it is potentially causing increased use of last reserve Gram-positive antibiotics such as linezolid, tigecycline, and daptomycin. Also, costly resources are spent on infection control measures to prevent even further spread of VRE at hospitals. Further studies are needed in an effort to differentiate the causes of the observed high mortality burden associated with nosocomial E. faecium BSI. Compatible with our overall null finding in national data, recent clinical data from one hospital suggest E. faecium is not the most likely cause of death [30].

Acknowledgements

We thank all Departments of Microbiology in Denmark for isolate submissions to NRL at SSI, and the collaboration on data reported to MiBa relevant for the present study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data are becoming or are already available for research upon reasonable request to Statens Serum Institut and within the framework of the Danish data protection legislation and any required permission from relevant authorities.